尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证正在加载完整 README。 以下为已保存的摘录。
A browsable archive of long-form animation in three sections: Donghua (Chinese), Anime (Japanese) and Korean (webtoon and manhwa adaptations), from official YouTube channels plus four English-dub channels, sorted into quality tiers. Every play button opens or embeds the YouTube video; nothing is downloaded or rehosted.查看完整 README ↗A browsable archive of long-form animation in four sections: Donghua (Chinese), Anime (Japanese), Korean (webtoon and manhwa adaptations) and Mature (adult-oriented, story-driven, never explicit), from official YouTube channels plus four English-dub channels, sorted into quality tiers. Every play button opens or embeds the YouTube video; nothing is downloaded or rehosted.
score() in engine/tiers.py (rating, length, upload gaps, anim). Series you have started come first. A slot only changes hands when you finish a series or mark it "Not for me". An original and its English-dub upload share one slot. Tabs: Top 15 (all), Top 15 Anime, Top 15 Donghua, each with its own slots.progress.json in this repo using a fine-grained GitHub token kept in the browser (this repo only, Contents read and write); the newer copy wins, so a cleared browser or another device gets everything back. progress.json sits outside engine/, so saving it does not rebuild the site. Back up / Restore to a file also works..github/workflows/nightly.yml): asks the YouTube Data API for uploads newer than the ones already saved, reads each official channel's members-only playlist ("UUMO" + channel ID) to mark locked uploads, checks new series' comments for AI slop, rebuilds index.html, publishes to GitHub Pages.engine/curated.json; that push triggers a rebuild.engine/ or the workflow rebuilds and republishes in about 2 minutes. Put [skip ci] in a commit message to push without publishing (the curation job does this for its progress commits).| Path | What |
|---|---|
engine/config.json | Channels: channels (English-dub, tier 3/6) and candidates (official, subtitled or dubbed with original sound; origin CN, JP or KR) |
engine/curated.json | Hand curation. meta[series_key]: name, rating, rating_src, completed, hide, quality, anim (animation quality 1 to 5; raises or lowers the series' score) with anim_note and anim_tried, members (hide as members-only by hand), members_ok (keep one the members-only detection got wrong), members_checked, merge_into with season_as (file the source as season N of the target), skip_seasons (season labels whose episodes another season already holds), from_ep (start the series at this episode; earlier ones are watched in another key), origin (CN, JP, KR), summary (English premise shown on the series page), summary_status ("unavailable" when researched but no reliable premise was found; the page then says so instead of "not written yet"), note, checked (date last researched), checked_eps (episode count then), verify_tried (date a recalled score could not be verified); the nightly run uses these to work only on deltas |
engine/videos.json, engine/scan/*.json | Saved uploads (one per line), with first_seen dates |
engine/scan/quality.json | AI verdict per checked video: quality, ai, slop, unknown |
engine/mature.py, engine/mature.json | Nightly AniList lookup: tags Seinen, Josei or gory series as Mature (anything AniList flags adult is skipped) and saves official "Also watch on" links; hand settings mature and watch in curated meta win |
engine/tiers.py | Score, tier rules, origin split, members-only hiding (see its docstring) |
engine/epparse.py | Title parser: series, season, episode; drops trailers, clips, members-only |
Secret needed: YT_API_KEY (YouTube Data API v3 key). Free quota is 10,000 units a day; a normal night uses under 500.